GWAS Strategy
GWAS was run using MLM model in GCTA1.93.2. Note that I tried different strategies to directly fit covariates or pre-adjust phenotypes by the covariates. The beta correlation between different strategies will be shown for Manhattan Plots as below.
Take SRS_RMB_sum in Probands (with FSIQ included as covariate) for example, I tried:
- Strategy 1: Phenotype pre-adjusted by age, sex, chip, FSIQ
- Strategy 2: Phenotype pre-adjusted by age, sex, chip, FSIQ and 20 PCs
- Strategy 3: directly fit age, sex, chip, FSIQ
- Strategy 4: directly fit age, sex, chip, FSIQ and 20 PCs
#grid.raster(readPNG("figures/beta_strategy.png")
grid.raster(readPNG("figures/beta_strategy.png"))
Note that considering the GWAS sample size, computational time and false positive rates, we will report the results below:
- based on Strategy2 for the same phenotype
- based on sum measurement for the same phenotype
Probands
All Individuals
- GWAS was run on all individuals including diverse ancestry backgrounds.
- Signals with association p-value < 1e-5 will be shown for Manhattan Plots.
Association Summary
Fitting FSIQ
datatable(iqs2, rownames = FALSE, filter="top", options = list(pageLength = 5, scrollX=T) )
Not fitting FSIQ
datatable(noiqs2, rownames = FALSE, filter="top", options = list(pageLength = 5, scrollX=T) )
QQ Plot
Primary Variable
Fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjIQ_primary.png")
grid.raster(img)

Not fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjnoIQ_primary.png")
grid.raster(img)

Secondary Variable
Fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjIQ_secondary.png")
grid.raster(img)

Not fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjnoIQ_secondary.png")
grid.raster(img)

Manhattan Plot
Primary Variable
Fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
grid.raster(img)

Not fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_noIQ_1e-5_primary_withPCs.png"))

Secondary Variable
Fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_secondary_withPCs.png"))

Not fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_noIQ_1e-5_secondary_withPCs.png"))

Europeans Only
- 6861972 QCd SNPs with MAF > 0.01 included
- 1946 European individuals are included
- Signals with association p-value < 1e-5 will be shown for Manhattan Plots.
Association Summary
Fitting FSIQ
datatable(iqs_EUR2, rownames = FALSE, filter="top", options = list(pageLength = 5, scrollX=T) )
Not fitting FSIQ
datatable(noiqs_EUR2, rownames = FALSE, filter="top", options = list(pageLength = 5, scrollX=T) )
QQ Plot
Primary Variable
Fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjIQ_primary_EUR.png")
grid.raster(img)

Not fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjnoIQ_primary_EUR.png")
grid.raster(img)

Secondary Variable
Fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjIQ_secondary_EUR.png")
grid.raster(img)

Not fitting FSIQ
#grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_withPCs.png")
img <- readPNG("figures/qqplot_probands_adjnoIQ_secondary_EUR.png")
grid.raster(img)

Manhattan Plot
Primary Variable
Fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_primary_EUR_withPCs.png"))

Not fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_noIQ_1e-5_primary_EUR_withPCs.png"))

Secondary Variable
Fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_adjIQ_1e-5_secondary_EUR_withPCs.png"))

Not fitting FSIQ
grid.raster(readPNG("figures/manhplot_probands_noIQ_1e-5_secondary_EUR_withPCs.png"))

Probands & Unaffected Siblings
- Combined phenotypes for Probands and Unaffected Siblings are pre-adjusted by covariates (20 PCs, sex, chip and ASD diagnosis) and then RINT.
- Based on the phenotype distribution, we run GWAS on combined data.
Note that there is no available information for FSIQ and Age for Unaffected Siblings, thus FSIQ and Age will not be included as covariates for the combined data analysis.
All Individuals
- GWAS was run on all individuals including diverse ancestry backgrounds.
- Signals with association p-value < 1e-5 will be shown for Manhattan Plots.
Association Summary
datatable(noiqs_probSibs2, rownames = FALSE, filter="top", options = list(pageLength = 5, scrollX=T) )
QQ Plot
Primary Variable
grid.raster(readPNG("figures/qqplot_probSibs_adjnoIQ_primary.png"))

Secondary Variable
grid.raster(readPNG("figures/qqplot_probSibs_adjnoIQ_secondary.png"))

Manhattan Plot
Primary Variable
grid.raster(readPNG("figures/manhplot_probandsAndUnaffSibs_noIQ_1e-5_primary_withPCs.png"))

Secondary Variable
grid.raster(readPNG("figures/manhplot_probandsAndUnaffSibs_noIQ_1e-5_secondary_withPCs.png"))

Europeans Only
- GWAS was run on European individuals with N=3544.
- Signals with association p-value < 1e-5 will be shown for Manhattan Plots for Manhattan Plots.
Association Summary
datatable(noiqs_probSibs_EUR2, rownames = FALSE, filter="top", options = list(pageLength = 5, scrollX=T) )
QQ Plot
Primary Variable
grid.raster(readPNG("figures/qqplot_probSibs_adjnoIQ_primary_EUR.png"))

Secondary Variable
grid.raster(readPNG("figures/qqplot_probSibs_adjnoIQ_secondary_EUR.png"))

Manhattan Plot
Primary Variable
grid.raster(readPNG("figures/manhplot_probandsAndUnaffSibs_noIQ_1e-5_primary_EUR_withPCs.png"))

Secondary Variable
grid.raster(readPNG("figures/manhplot_probandsAndUnaffSibs_noIQ_1e-5_secondary_EUR_withPCs.png"))

Heritability Estimation
Probands
- Unrelated Individuals (GRM < 0.05) were used for performing GREML analysis.
- Using different models based on phenotypes:
- Est.1-1: pre-adjust by Age, Sex and Chip
- Est.1-2: directly fit covariates: Age, Sex and Chip
- Est.2-1: pre-adjust by Age, Sex, Chip and 20PCs
- Est.2-2: directly fit covariates: Age, Sex, Chip and 20PCs
- Est.3-1: pre-adjust by Age, Sex, Chip and FSIQ
- Est.3-2: directly fit covariates: Age, Sex, Chip and FSIQ
- Est.4-1: pre-adjust by Age, Sex, Chip, FSIQ and 20PCs
- Est.4-2: directly fit covariates: Age, Sex, Chip, FSIQ and 20PCs
Primary Variable
grid.raster(readPNG("figures/h2_probands_primary.png"))

Secondary Variable
grid.raster(readPNG("figures/h2_probands_secondary.png"))

Probands & Unaffected Siblings
- Unrelated Individuals (GRM < 0.05) were used for performing GREML analysis.
- Using different models based on phenotypes:
- Est.1-1: pre-adjust by ASD diagnosis, Sex and Chip
- Est.1-2: directly fit covariates: ASD diagnosis, Sex and Chip
- Est.2-1: pre-adjust by ASD diagnosis, Sex, Chip and 20PCs
- Est.2-2: directly fit covariates: ASD diagnosis, Sex, Chip and 20PCs
Primary Variable
grid.raster(readPNG("figures/h2_probSibs_primary.png"))

Secondary Variable
grid.raster(readPNG("figures/h2_probSibs_secondary.png"))

Genetic Correlation
Probands
RRBs
RRBs VS. Public
Probands & Unaffected Siblings
RRBs